Author + information
- Received July 30, 2003
- Revision received October 30, 2003
- Accepted November 3, 2003
- Published online April 7, 2004.
- Srinivas Dukkipati, MD*,
- William W O'Neill, MD, FACC*,
- Kishore J Harjai, MD, FACC*,
- William P Sanders, MD†,
- Datinder Deo, MD*,
- Judith A Boura, MS*,
- Beth A Bartholomew, MD*,
- Michael W Yerkey, MD*,
- H.Mehrdad Sadeghi, MD* and
- Joel K Kahn, MD, FACC*,* ()
- ↵*Reprint requests and correspondence:
Dr. Joel K. Kahn, William Beaumont Hospital, 3601 West Thirteen Mile Road, Royal Oak, Michigan 48073, USA.
Objectives We sought to identify the incidence, predictors, and clinical implications of cerebrovascular accidents (CVAs) after percutaneous coronary interventions (PCIs).
Background Cerebrovascular accidents after PCI, although rare, can be devastating. Limited information exists regarding the characterization of this complication.
Methods The study population comprised 20,679 patients who underwent PCI between September 1993 and April 2002. A CVA was defined as a composite of transient ischemic attack (TIA) and stroke. The characteristics of those who had a periprocedural CVA were compared with those who did not.
Results A CVA occurred in 92 patients (0.30% of procedures). Of these, TIA occurred in 13 patients (0.04%) and stroke in 79 patients (0.25%). On multivariate analysis, patients with this complication more frequently had diabetes mellitus (adjusted odds ratio [OR] 1.8, 95% confidence interval [CI] 1.1 to 3.0; p = 0.013), hypertension (OR 1.9, 95% CI 1.1 to 3.3; p = 0.033), previous CVA (OR 2.3, 95% CI 1.3 to 4.0; p = 0.0059), and creatinine clearance ≤40 ml/min (OR 3.1, 95% CI 1.8 to 5.2; p < 0.0001). They underwent urgent or emergent procedures (OR 2.7, 95% CI 1.3 to 5.5; p = 0.0092) with more thrombolytic (OR 4.7, 95% CI 2.3 to 9.7; p < 0.0001) and intravenous heparin (OR 1.9, 95% CI 1.1 to 3.4; p = 0.030) use before PCI, and they more often required emergent intra-aortic balloon pump placement (OR 2.2, 95% CI 1.1 to 4.3; p = 0.028). On multivariate analysis, CVA was independently associated with in-hospital death (OR 7.8, 95% CI 4.2 to 14.7; p < 0.0001), acute renal failure (OR 2.8, 95% CI 1.4 to 5.7; p = 0.0042), and new dialysis (OR 3.73, 95% CI 1.01 to 13.8; p = 0.049) after PCI.
Conclusions Cerebrovascular accidents after PCI, although rare, are associated with high rates of in-hospital death and acute renal failure, often requiring dialysis.
Acute cerebrovascular accidents (CVAs) after percutaneous coronary interventions (PCIs), although rare, are associated with high rates of mortality and morbidity (1–6). Although previous studies have been predominantly limited to the characterization of stroke after PCI, few have included transient ischemic attacks (TIAs) in their analyses (6). The incidence of these combined cerebrovascular complications has been reported to range from 0.27% to 0.50% (1,2). The low incidence of these complications has limited the ability of previous studies to adequately characterize stroke and TIA after PCI. Therefore, we sought to identify the incidence, predictors, and in-hospital outcomes of these combined complications in 20,679 consecutive patients who underwent PCI at a single, large-volume center.
Definition of CVA
A CVA was defined as the onset of a new neurologic deficit that occurred anytime after PCI during the index hospitalization. If the duration of the deficit was <24 h, it was defined as a TIA. If the deficit persisted for a longer period, it was defined as a stroke. The diagnosis of CVA was made by experienced neurologists.
The clinical, angiographic, and in-hospital outcomes of all patients who had PCI from September 1993 through April 2002 at the William Beaumont Hospital were stored prospectively in a database. The 334 PCI patients who also underwent coronary artery bypass grafting during the index hospitalization were excluded. None of the excluded patients suffered a CVA before surgery. The remaining 20,679 patients underwent 30,998 PCI procedures during this time period. In patients with multiple procedures, only data from the last visit were used in subsequent analyses. However, if a CVA occurred in a hospital stay other than the last one, then data from that visit were used instead. The clinical and angiographic features, as well as in-hospital outcomes of the patients who suffered an in-hospital CVA after PCI, were compared with those who did not.
A retrospective chart review of all CVA patients was performed to obtain information that was unavailable in the computer database. The information collected in this manner consisted of presenting with neurologic deficits, time of onset after PCI, neurologists' clinical diagnosis, type and location of CVA, duration of deficits, and disposition at hospital discharge. Most patients had computed tomography or magnetic resonance imaging studies of the brain performed and interpreted at the time of hospitalization by independent neuroradiologists. Based on their reports, another experienced neuroradiologist retrospectively classified the stroke type (ischemic or hemorrhagic) and location (e.g., frontal, parietal). If the findings reported by the independent neuroradiologists were questionable or negative, the imaging studies were reviewed again for confirmation and classification of CVA.
The baseline creatinine clearance was calculated using the Cockcroft-Gault equation (7). Urgent PCI was defined as PCI in hemodynamically stable patients during hospitalization for unstable angina or non–Q-wave myocardial infarction (NQMI). Emergent PCI was defined as immediate PCI for acute ST-segment elevation myocardial infarction (MI) or a hemodynamically unstable acute coronary syndrome. Post-PCI NQMI was defined as the presence of two of the following three criteria: prolonged chest pain, elevation of serum creatine kinase ≥3 times the upper limit of normal, or electrocardiographic (ECG) changes suggesting ischemia. Q-wave MI after PCI was defined as the presence of two of the three aforementioned criteria and the appearance of new pathologic Q waves on ≥2 contiguous ECG leads. Acute renal failure after PCI was defined as ≥1 mg/dl elevation in serum creatinine above the baseline value before PCI.
All statistical analyses were performed using SAS version 8.0 software (Cary, North Carolina). Continuous variables are expressed as the mean value ± SD. Comparisons between groups were performed using the Student ttest. Categorical variables are expressed as counts and percent frequencies and were compared using the chi-square test. Comparisons between the three types of CVA were made using a Kruskal-Wallis test for continuous variables and a chi-square test for categorical variables, if appropriate (expected frequency >5). Otherwise, the Fisher exact test was used. Step-down multivariate logistic regression analysis was performed to identify independent predictors of in-hospital CVA. Variables with a univariate relation (p ≤ 0.05) with CVA or thought to be important were included in the model. The Cstatistic was used to determine the discriminatory power of the model. The Cstatistic is a rank-correlation index that measures the association between predicted probabilities and observed responses. A perfect correlation would be 1. The adjusted odds ratio (OR) and 95% confidence interval (CI) were calculated for each variable in the final model. Furthermore, we performed multiple logistic regression to assess the independent impact of in-hospital CVA on clinical outcomes (death, renal insufficiency, and new dialysis).
Baseline patient characteristics
In-hospital CVA after PCI occurred in 92 of 20,679 patients (0.30% of procedures). Of these, a TIA occurred in 13 (0.04%) patients and stroke in 79 (0.25%) patients. The incidence of CVA after PCI was significantly higher in the period before 1998 compared with 1998 and after (0.39% vs. 0.22%, p = 0.0075). The baseline clinical differences between patients who had a periprocedural CVA and those who did not are shown in Table 1. Patients with a CVA were older females with a smaller body surface area. Those with CVA more often had a history of hypertension, diabetes mellitus, congestive heart failure, previous CVA, peripheral vascular disease, and worse baseline renal function, as measured by creatinine clearance. Cardiac catheterization was more often urgent or emergent in the CVA group, with a significantly greater prior use of intravenous heparin and thrombolytics.
Data on PCI
As shown in Table 2, patients who had a CVA underwent longer cardiac catheterization procedures with more iodinated contrast. Saphenous vein graft PCI was more frequent (15% vs. 7%, p = 0.003) with a trend toward more left main artery interventions (4.4% vs. 1.8%, p = 0.09) in the CVA group. Endotracheal intubation and an intra-aortic balloon pump (IABP) were needed in 11% and 20% of this group, respectively. However, there were no significant differences in the rates of no reflow, coronary perforation, or abrupt closure. No significant differences were observed in the maximum activated clotting time or glycoprotein IIb/IIIa use.
Clinical presentation and CVA characteristics
The most common presenting neurologic manifestations in patients with CVA were motor or speech deficits (Table 3). The combination of unresponsiveness or altered mental status occurred in 45% of all CVA patients and together represented the most frequently reported deficits. Of the 92 patients who were diagnosed with a CVA, 4 patients had severe internal carotid artery stenosis that was diagnosed after the onset of symptoms. All four of them were diagnosed by carotid Doppler ultrasound examinations. Three of these patients had a TIA and one had a stroke. In all cases, the presenting symptoms could be localized to the ipsilateral hemisphere as the stenosis. Fifty-seven (62%) patients manifested neurologic deficits within the first 24 h after PCI. Seventeen (18%) patients presented between 24 and 48 h and 18 patients (20%) presented after 48 h following PCI.
Thirteen patients had a TIA and 79 had a stroke (Fig. 1). Of the 12 TIA patients who underwent neuroimaging, 2 had acute ischemic infarcts (1 in the distribution of the middle cerebral artery and 1 in the distribution of the posterior cerebral artery), and the other 10 had no imaging abnormalities. Delayed computed tomographic scans were not routinely performed and may have identified a greater number of patients with infarcts. In the 71 stroke patients who underwent an imaging study, the infarct location could be attributed to the following artery distributions: middle cerebral in 56%, anterior cerebral in 2%, posterior cerebral in 37%, superior cerebellar in 5%, posterior inferior cerebellar in 5%, and basilar in 7%. Infarcts to the anterior circulation represented 58% and those to the posterior circulation represented 54% of the infarcts. Three of the 43 patients (7%) with stroke and evidence of infarct on neuroimaging had infarcts involving more than one artery distribution. Hemorrhage occurred in the following locations: temporal-parietal-occipital lobes in 36%, frontal lobe in 27%, basal ganglia in 9%, intraventricular hemorrhage in 46%, subdural hemorrhage in 36%, epidural hemorrhage in 18%, and subarachnoid hemorrhage in 46%.
Death occurred in 25% of those with CVA, compared with 1.5% of those without CVA (p < 0.0001). None of the deaths occurred in those with a TIA. However, 69% of patients with hemorrhagic stroke and 21% of patients with ischemic stroke died in the hospital. Patients with CVA had a higher incidence of acute renal failure (15% vs. 1.6%, p < 0.0001) and new dialysis (4.4% vs. 0.3%, p < 0.0001), compared with those who did not suffer this complication. Although recurrent angina within the first 24 h after PCI was more frequent in the CVA group (7.6% vs. 2.4%, p = 0.0072), there were no differences in the rates of NQMI (1.1% vs. 0.7%, p = 0.46) and Q-wave MI (0% vs. 0.3%, p = 1.00). A CVA was also associated with a longer length of hospital stay (10.3 ± 8.7 vs. 2.4 ± 4.5 days, p < 0.0001).
Multivariate predictors of CVA
Independent predictors of in-hospital CVA, with their respective adjusted OR and 95% CI, are shown in Figure 2. Included in the multivariate analysis were age >70 years, gender, body surface area, diabetes mellitus, hypertension, hypercholesterolemia, congestive heart failure, previous CVA or PCI, creatinine clearance ≤40 ml/min, peripheral vascular disease, urgent or emergent catheterization, thrombolytic or heparin use before PCI, fluoroscopy time, contrast amount, coronary perforations, no reflow, intervention to saphenous vein graft, planned or unplanned IABP use, and year prior to 1998. As there were changes in anticoagulation strategies with increasing glycoprotein IIb/IIIa use and decreasing postprocedural heparin use around 1998, this year was used to divide the study population. The Cstatistic for the model was 0.80, illustrating good discriminatory power. All of these variables and in-hospital CVA were used to determine the independent predictors of in-hospital death, acute renal failure, and new dialysis after PCI. A CVA was independently associated with in-hospital death (OR 7.8, 95% CI 4.2 to 14.7; p < 0.0001), acute renal failure (OR 2.8, 95% CI 1.4 to 5.7; p = 0.0042), and new dialysis (OR 3.73, 95% CI 1.01 to 13.8; p = 0.049). The Cstatistic was 0.91 for death, 0.92 for acute renal failure, and 0.94 for new dialysis.
Differences based on type of CVA
Most of the baseline clinical and cardiac intervention characteristics were not different between those who had a hemorrhagic stroke, ischemic stroke, or TIA. The rates of heparin use before PCI and glycoprotein IIb/IIIa use were not significantly different between the three groups. There was a borderline significantly higher use of thrombolytics before PCI (38% vs. 12% vs. 8%, p = 0.060) and contrast amount (322 ± 127 ml vs. 252 ± 134 ml vs. 246 ± 77 ml, p = 0.085) in those with hemorrhagic stroke, compared with those with ischemic stroke or TIA. Although there were no differences in the rates of acute renal failure or new dialysis, there were significantly more deaths in the hemorrhagic stroke group than in the ischemic stroke or TIA groups (69% vs. 21% vs. 0%, p = 0.0001).
Of 69 patients with CVA who survived the hospitalization, 72% had a persistent neurologic deficit at the time of hospital discharge. All of these deficits were in patients with ischemic or hemorrhagic strokes and not in TIA sufferers. Of these 69 patients, most were discharged home (43%) without any further care needed. However, skilled home care was needed in 26% of patients, nursing home or assisted living in 9%, and in-patient rehabilitation in 22%.
Our study is the largest observational analysis of the incidence, predictors, and clinical implications of post-PCI CVA. In approximately 1 of 300 procedures, CVA complicates PCI. Patients with this complication had higher rates of hypertension, diabetes mellitus, previous CVA, and impaired renal function. They more frequently had cardiac catheterization for urgent or emergent reasons, with a greater prior use of thrombolytics and intravenous heparin. These patients underwent longer cardiac catheterization procedures with a greater use of contrast. Cerebrovascular events after PCI were associated with high rates of in-hospital complications, including an eight-fold adjusted risk of death, a three-fold risk of acute renal failure, and a four-fold risk of requiring dialysis. All of the deaths occurred in patients with an ischemic or hemorrhagic stroke, and TIA was associated with 0% mortality. However, the rates of acute renal failure and new dialysis were the same in all three groups. In the patients who survived the hospitalization, 72% had a persistent deficit at the time of discharge. All of the persistent deficits were once again limited to those with an ischemic or hemorrhagic stroke.
The incidence of TIA and stroke in our study is within the range reported in other studies (1,2). We found that hypertension, diabetes mellitus, impaired renal function, and a history of CVA are independently associated with periprocedural CVA. Hypertension and diabetes mellitus are both well-established risk factors for stroke (8). Although hypertension, diabetes mellitus, chronic renal dysfunction, and a history of CVA have been well established as risk factors for periprocedural stroke in patients undergoing coronary artery bypass grafting (9), they have not been reported in previous PCI studies (1,4–6,10,11). Rubenstein et al. (12)showed a borderline significant excess of stroke in those with baseline renal dysfunction, as compared with control subjects, after PCI. Patients with chronic renal insufficiency often have concomitant cardiovascular and cerebrovascular risk factors, such as diabetes mellitus and hypertension. In our analysis, even after adjusting for these risk factors, chronic renal insufficiency conferred an additional risk. The mechanism of stroke in these patients with advanced chronic renal insufficiency may be related to platelet dysfunction and bleeding diathesis, often present in this group (13).
We found that thrombolytic and intravenous heparin use before PCI was independently associated with periprocedural CVA. Thrombolytic therapy in the setting of an acute MI is a well-established risk factor for both hemorrhagic and nonhemorrhagic stroke (14,15). Even intracoronary thrombolytic therapy has been described as an independent risk factor for hemorrhagic stroke (16). However, the association between intravenous heparin and stroke is less clear. Although previous studies have found a borderline significant association, the observed differences are thought to be inconclusive because of the low number of patients involved in the studies (17–19).
After adjusting for prior treatment with thrombolytics and intravenous heparin, urgent or emergent catheterization conferred an additional risk for periprocedural CVA. Possible explanations for this include the greater propensity for hemodynamic compromise in these patients, which may increase the risk of ischemic stroke, and the less meticulous care in advancement of catheters through the aorta during urgent or emergent PCI, which increases the risk of scraping aortic plaque and increases the risk of embolization to the brain. Keeley and Grines (20)showed that scraping of aortic plaque occurs in more than 50% of percutaneous revascularization procedures. Furthermore, Eggebrecht et al. (21)showed that larger-caliber catheters scrape more aortic plaque than smaller catheters. Therefore, the increased risk of CVA after urgent or emergent procedures may reflect the scraping of more aortic plaque in these cases, thereby increasing the risk of emboli to the brain. Additionally, those with CVA also underwent longer procedures, thereby further increasing the likelihood of aortic plaque embolization. During cardiac catheterization, the amount of contrast agent used in those with CVA was significantly higher and was independently associated with this complication. This may relate to the previously described thrombogenic potential of some types of contrast agents (22). The association between IABP and periprocedural stroke has previously been reported (1,23).
We noted a temporal decrease in CVA when comparing interventions from 1998 to 2002 with those performed from 1993 to 1997 (0.22% vs. 0.39%, p = 0.0075). This trend may reflect improvement in instrumentation and techniques involved with coronary interventions over the years. Both arterial sheath and catheter sizes used in coronary interventions have decreased. Smaller catheters disrupt less aortic plaque, thereby diminishing the likelihood of embolization to the brain (20,21). Furthermore, heparin use after PCI has decreased, particularly after 1998, due to evidence for a lack of its benefit (24,25). These factors may have contributed to the observed temporal trend in CVA.
The increased rates of mortality and morbidity in patients with cerebrovascular complications after PCI are well known (1,6). However, the finding of high rates of acute renal failure and new dialysis in patients with periprocedural CVA has not previously been reported. After accounting for other baseline characteristics, including a history of baseline renal insufficiency and amount of contrast agent used, periprocedural CVA is independently associated with these adverse outcomes. Perhaps, periprocedural CVA is a marker of systemic embolization that occurs during cardiac catheterization, another manifestation of which is acute renal failure and possibly dialysis. In our study, patients who had a CVA were older and had higher rates of hypertension and peripheral vascular disease, which are both risk factors for atherosclerotic aortic plaque (26). Individuals with a higher atherosclerotic burden in the aorta have higher rates of systemic embolization, either spontaneously or from catheter manipulation (26–28). In an autopsy series of 29 patients who had cholesterol emboli to the brain, 55% also had emboli to their kidneys (29). The majority of the patients in this series underwent procedures involving manipulation of the aorta, and 10% of patients died due to complications of acute renal failure.
Few recommendations can be made regarding the prevention of CVA after PCI, as most risk factors are not modifiable, including a history of advanced renal insufficiency, diabetes mellitus, hypertension, and previous CVA. However, in patients with these risk factors who present with an acute MI, primary angioplasty should be favored over thrombolytics. During cardiac catheterization, careful attention should be made to minimize catheter manipulation and exchanges. The length of the procedure and amount of contrast used should also be minimized. Postprocedural anticoagulation with heparin should be eliminated in patients with uncomplicated coronary interventions. In patients who present with focal neurologic deficits soon after PCI, consideration should be given to immediate neuroimaging and percutaneous cerebral intervention.
The retrospective, single-institution methodology and the absence of long-term follow-up are limitations of this study. Because the incidence of periprocedural CVA is very low, the study does not have the power to discriminate differences between TIA, ischemic stroke, and hemorrhagic stroke. The incidence of TIA is likely higher than that reported in our study, as transient neurologic deficits may not have come to the attention of physicians or may not have been identified. We recognize that the pathophysiology of CVA that occurred >48 h after the coronary intervention may not necessarily be related to the PCI. However, because a change in mental status was one of the presenting symptoms in 32% of CVA patients, this may have led to a delayed diagnosis. Furthermore, postprocedural anticoagulation may have contributed to delayed hemorrhage in CVA patients. Because of these factors, we have chosen to include all patients, regardless of the time frame of presentation.
Cerebrovascular accidents complicating PCI are rare and have decreased in incidence as interventional techniques and instrumentation have improved. Many of the risk factors associated with this complication are not modifiable. With primary angioplasty replacing thrombolytics as the treatment of choice for acute MI, minimization of prolonged postprocedural anticoagulation, as well as continued improvement in interventional technology and techniques, we will likely continue to reduce this devastating complication.
We thank Sue Glazier, Linda Mizejewski, and Linda McWilliams for their invaluable assistance in gathering and organizing the necessary patient information needed for the completion of this study. We also thank Sue Tomaszycki for her assistance with the illustrations in the manuscript.
This study was presented in part as an abstract at the American College of Cardiology Scientific Sessions in Chicago, Illinois, in March 2003.
- confidence interval
- cerebrovascular accident
- intra-aortic balloon pump
- myocardial infarction
- non–Q-wave myocardial infarction
- odds ratio
- percutaneous coronary intervention
- transient ischemic attack
- Received July 30, 2003.
- Revision received October 30, 2003.
- Accepted November 3, 2003.
- American College of Cardiology Foundation
- Fuchs S.,
- Stabile E.,
- Kinnaird T.D.,
- et al.
- Dorros G.,
- Cowley M.J.,
- Simpson J.,
- et al.
- Bredlau C.E.,
- Roubin G.S.,
- Leimgruber P.P.,
- Douglas J.S.,
- King S.B. III.,
- Gruentzig A.R.
- Goldstein L.B.,
- Adams R.,
- Becker K.,
- et al.
- Stamou S.C.,
- Hill P.C.,
- Dangas G.,
- et al.
- Szczech L.A.,
- Best P.J.,
- Crowley E.,
- et al.
- Rubenstein M.H.,
- Harrell L.C.,
- Sheynberg B.V.,
- Schunkert H.,
- Bazari H.,
- Palacios I.F.
- Mahaffey K.W.,
- Granger C.B.,
- Sloan M.A.,
- et al.
- Collins R.,
- MacMahon S.,
- Flather M.,
- et al.
- Keeley E.C.,
- Grines C.L.
- Stone G.W.,
- Marsalese D.,
- Brodie B.R.,
- et al.
- Rabah M.,
- Mason D.,
- Muller D.W.M.,
- et al.
- Tunick P.A.,
- Rosenzweig B.P.,
- Katz E.S.,
- Freedberg R.S.,
- Perez J.L.,
- Kronzon I.